Vectorizing integrate()
I found mixed (and not always easy to predict) results from the byte-code compiler. It seems necessary to test whether it helps. On some calculations, it is definitely worthwhile. JN
On 12-12-07 01:57 PM, Berend Hasselman wrote:
On 07-12-2012, at 19:37, Spencer Graves wrote:
On 12/7/2012 9:40 AM, Berend Hasselman wrote:
benchmark(eta1 <- f1(X, B, x, sem1), eta2 <- f2(X, B, x, sem1), eta3 <- f3(X, B, x, sem1),
+ eta4 <- f4(X, B, x, sem1), eta5 <- f5(X, B, x, sem1), eta6 <- f6(X, B, x, sem1),
+ replications=10, columns=c("test","elapsed","relative"))
test elapsed relative
1 eta1 <- f1(X, B, x, sem1) 1.873 1.207
2 eta2 <- f2(X, B, x, sem1) 1.552 1.000
3 eta3 <- f3(X, B, x, sem1) 1.807 1.164
4 eta4 <- f4(X, B, x, sem1) 1.841 1.186
5 eta5 <- f5(X, B, x, sem1) 1.852 1.193
6 eta6 <- f6(X, B, x, sem1) 1.601 1.032
As you can see using the compiler package is beneficial speedwise.
f2 and f6, both the the result of using the compiler package, are the quickest.
It's quite likely that more can be eked out of this.
So the compiler (f2, f4, f6) provided a slight improvement over f1 and f3 but not f2, and in any event, the improvement was not great.
I don't understand the "but not f2". And I don't understand the conclusion for (f2,f4,f6). f4 is a compiled version of f3 and is slower than its non compiled version. f2 and f6 are the quickest compiled versions. Indeed the improvement is not earth shattering but it does demonstrate what you can achieve by using the compiler package. Berend